CCNCS Seminar Details
Do familiarity and recollection originate from the same memory trace? A computational memory model
|Speaker:||Dr Andrea Greve|
|Date/Time:||Wednesday 8 April 2009, 4.15pm|
|Location:||SB110B, Computing Laboratory, Cornwallis|
Dual process theories of episodic memory state that retrieval is contingent upon two independent processes: familiarity (providing a sense of oldness) and recollection (the recovery of events and context). Neuroimaging studies have reported distinct neural signatures for familiarity and recollection, lending further support to dual-process theory. However, it remains unclear whether these signatures reflect activation of distinct memory traces or different retrieval mechanisms operating on a single memory trace. This question is important because the competing explanations have very different implications. The first option necessitates storage of two distinct memory traces, which further requires accounting for how these traces are encoded and linked at retrieval. By contrast, the second alternative does not demand distinct traces. To date it is unknown whether empirical findings can be accounted for by distinct retrieval processes accessing a single memory trace. In my talk I will present a computational model that uses a single neuronal network to store memory traces, with two distinct and independent retrieval processes accessing the memory store. The model is capable of discriminating old from new items after a single exposure, and exhibits ROC curves analogous to those found in empirical data. The putative familiarity and recollection processes exhibit distinct characteristics in the model, for example, differing in capacity and sensitivity to sparse and convoluted patterns. By demonstrating that a dual-process single-trace model can account for a range of empirical findings, I will highlight the importance of distinguishing between neuronal processes and the neuronal representations upon which they operate.